Re: Compiling CVS HEAD with clang under OSX

Neil Conway <neil(dot)conway(at)gmail(dot)com> writes:
> (As an aside, is "no-cpp-precomp" still necessary for
> reasonably-modern versions of Apple GCC?)

I looked into this point a little bit. Apple abandoned their
nonstandard precompiler as of gcc 3.3, so the switch is a no-op
in that version and later, as per release notes here:
http://developer.apple.com/legacy/mac/library/releasenotes/DeveloperTools/RN-GCC3/index.html

I have a copy of gcc-3.3 still in captivity, and have verified that it
builds 9.0beta4 just fine without the no-cpp-precomp switch.

So the question is whether anybody is likely to still be using
even-older versions of Apple's gcc. I think probably not: as of Xcode
2.0, released in early 2005, 3.3 was already the oldest supported branch
AFAICT. If you'd like to build code that will run on Intel Macs, you
have to be using gcc 4.0 or later, so even 3.3 is pretty much in the
dustbin of history.

So I think we might as well drop -no-cpp-precomp. If there is anyone
out there who really needs it, they can always push it in via a manual
setting of CPPFLAGS, but there seems little reason to include it by
default.

I'm still wondering about the bleats I saw for -fwrapv though.
configure already is set up to install that switch only conditionally:

# Disable optimizations that assume no overflow; needed for gcc 4.3+
PGAC_PROG_CC_CFLAGS_OPT([-fwrapv])

but apparently the test used for this does not notice warning messages.
Can we improve that?

regards, tom lane